Render time22.9s + 29.1s (two halves, final renders)
Format×720 · 16:9 landscape
Published2026-10-25
What broke (and the fix)
Nothing broke. The Opus-style headline spans nearly the full frame width — checked at full size, it does not clip. Both halves carry the NOT A MODEL BENCHMARK footer on screen, and the combined film was frame-checked in both halves.
Original promptTested
Same showreel prompt, two model styles — Opus 5.5 vs GPT-6 Astra — HyperFrames (format comparison, not a benchmark)
Banned list —Beat grid —Build method —Self-review —
<direction>
An 8-second landscape showreel for a showreel, in two formats. Opus-style half (label burned in: OPUS-STYLE PROMPT (AS WRITTEN)): headline THE SHOWREEL, AS WRITTEN. plus three kinetic words — BOLD / FAST / TRUE — plus the promise line THREE WORDS. ONE PROMISE. Astra-style half (label burned in: ASTRA-STYLE ADAPTATION): the same subject as three scene cards — OPENING / MIDDLE / FINALE — plus the line SCENES, NOT SLOGANS. Both halves carry the footer RENDERED BY OUR PIPELINE — NOT A MODEL BENCHMARK. Accent throughout: #ff5c1a. House style: near-black #0a0a0a background, white type, ONE orange accent #ff5c1a, Arial/Helvetica bold, no webfonts. Fictional brands/data only.
Banned: glows, gradients, particle bursts, and any claim that either half is a model's output or a benchmark result.
</direction>
<structure>
Each half 8s: label strip top, headline, then words (Opus-style) or scene cards (Astra-style), then the closing line. The two halves are concatenated as sequential halves with ffmpeg — Opus-style first, Astra-style second.
</structure>
<build>
Two HyperFrames + GSAP compositions, 1280x720 landscape, 30fps, seek-driven timelines. The Opus-style half uses kinetic-word staggers; the Astra-style half uses card staggers.
</build>
<gotchas>
- The disclosure is part of the film (footer on both halves) and part of this prompt — never ship the comparison without it.
- The Opus-style headline spans nearly the full frame width at 80px; verify at full size that it does not clip.
</gotchas>
<methods>
Methods note — DISCLOSURE, read this first: BOTH films were rendered by our own pipeline — one from the prompt as written for Opus, one adapted to the prompt structure Astra's creator published; this is a FORMAT comparison, NOT a live benchmark of the two models. We did not run Opus 5.5 or GPT-6 Astra, and no claim in this entry is a benchmark result. Final render times: Opus-style 22.9s + Astra-style 29.1s. Iterations: 1 per half, 2 total.
</methods>
Before you finish: step through the timeline at five points and fix anything that overlaps, clips the frame, or drifts off the beat grid.
DISCLOSURE: BOTH films were rendered by our own pipeline — one from the prompt as written for Opus, one adapted to the prompt structure Astra's creator published; this is a FORMAT comparison, NOT a live benchmark of the two models. We did not run Opus 5.5 or GPT-6 Astra, and nothing here is a benchmark result. The format difference itself is real and visible: the as-written half sells with words (BOLD / FAST / TRUE), the adapted half sells with scenes (OPENING / MIDDLE / FINALE). Render times: Opus-style 22.9s + Astra-style 29.1s; 1 iteration per half.